Search Results (277)

Federated learning (FL) enables collaborative model training across decentralized clients while preserving data privacy, yet real-world deployments still face communication bottlenecks, performance degradation under heterogeneous client data, and limited personalization. In this study, we introduce FedMCP++, a modular and communication-efficient personalized FL framework in which every client owns a complete private model—a lightweight convolutional backbone with a private expert head—and collaboration is carried out entirely through knowledge exchange rather than parameter exchange. In each round, clients share only temperature-softened class predictions and class-wise feature prototypes computed on a small public proxy set; the server fuses them into an accuracy-weighted teacher and broadcasts the result, and clients realign their models through knowledge distillation, an instance-level contrastive objective, and prototype alignment. We evaluate FedMCP++, its ablations, and two knowledge-based baselines on six benchmark vision datasets with 10, 20, and 30 clients. The results indicate dataset-dependent trade-offs rather than uniform superiority: collaborative distillation improves average client-level accuracy over independent local training in twelve of eighteen configurations—most clearly under severe per-client data scarcity (e.g., up to +2.7 percentage points on KMNIST and +2.4 on STL-10 with 20–30 clients)—whereas independent training ensembles remain strongest on SVHN and CIFAR-10 at the studied budgets. Because no parameters are transmitted, the per-round uplink payload is a fixed-size 42.6 KB message, 9.9–12.8× smaller than full-model synchronization, and is invariant to model capacity. These properties make FedMCP++ a flexible framework for personalized FL in wireless edge and Internet of Things environments where bandwidth and privacy constraints are paramount. Full article

Enhancing the proof mass without enlarging the chip area or introducing structural asymmetry is a central challenge in the development of low-noise microelectromechanical system (MEMS) accelerometers. Here, we present a symmetric triple-layer MOEMS accelerometer and analyze its device-level sensitivity trade-off, cross-axis coupling behavior, and dynamic consistency between measurement and finite-element simulations. The proposed sensing element sandwiches one without-beam mass layer between two identical with-beam layers, thereby increasing the effective proof mass while preserving mirror symmetry. A lumped-parameter model is developed to explain the sensitivity trade-off among single-layer, asymmetric double-layer, and symmetric triple-layer configurations. Finite-element simulations are used to distinguish translational cross-axis coupling from rotational cross-axis coupling. The experimental characterization of one packaged triple-layer prototype demonstrates a mechanical sensitivity of 193.91 µm/(m/s²), a 10 min output RMS fluctuation of 1.81 µg, and a measured first-order resonant frequency of 11.23 Hz, in close agreement with the tolerance-included finite-element prediction of 11.40 Hz. The resonance bandwidth further yields an apparent package-level quality factor of approximately 374 under ambient pressure, providing additional characterization of the packaged device dynamics. Full article

The growing demand for virtual reality and immersive applications has increased interest in 360° video streaming. When viewing omnidirectional content through a head-mounted display, users observe only a limited portion of the content, i.e., the viewport, at any given time. Consequently, transmitting the complete panoramic frame at uniformly high quality is bandwidth-inefficient. This review presents a system-level analysis of viewport-adaptive three-degree-of-freedom (3DoF) 360° video streaming, focusing on the coupled roles of viewport prediction, tile-based multi-rate encoding and bitrate allocation, transport mechanisms, and edge-assisted processing. The reviewed literature is examined to identify the design dependencies and trade-offs among these components. Viewport-adaptive approaches seek to reduce the bandwidth allocated to regions outside the instantaneous viewport while preserving the quality of the visible region. The analysis shows that their effectiveness cannot be attributed to prediction accuracy alone: the resulting Quality of Experience (QoE) depends jointly on tile granularity, bitrate allocation, buffer occupancy, transport delay, and whether prioritized tiles arrive before their playback deadlines. Finer tiling can improve spatial selectivity but increases coding, signaling, and request overhead. Moreover, HTTP/2, HTTP/3/QUIC, RTP/RTSP, and WebRTC present different reliability, latency, congestion-control, and scalability trade-offs across buffered video-on-demand, low-latency live streaming, and interactive immersive applications. Based on this synthesis, the review formulates a unified closed-loop cross-layer framework that coordinates prediction, tiling, bitrate allocation, request timing, transport configuration, buffering, and edge processing under bandwidth, latency, and resource constraints. Full article

Broadband, low-frequency sound absorption within a compact device remains a central unsolved problem in noise control engineering, arising from fundamental trade-offs among resonator volume, absorption bandwidth, panel thickness, and frequency tunability that no passive, linear, time-invariant system can simultaneously circumvent. This review establishes a unified mass–spring–damping (MSD) framework applied systematically across the full spectrum of resonance-based absorber technologies. From first principles, we derive the mass–stiffness coupling result (the mass-disappearing result of Shen and Liu): fixing the resonance frequency imposes $K=M{\omega }_{\mathrm{res}}^2$, so acoustic mass and stiffness cannot be adjusted independently; the half-absorption bandwidth ${\Pi }_1=\eta /\left(M{\omega }_{\mathrm{res}}\right)+V{\omega }_{\mathrm{res}}/\left(c_0{S}_{\mathrm{tar}}\right)$ then depends explicitly on the cavity volume V (system stiffness) and on the damping coefficient $\eta$, rather than on mass as an independent lever. This explains why neck extension, space-coiling, and membrane loading—which merely add mass while leaving the cavity stiffness unchanged—fail to broaden the absorption band at fixed volume, and refocuses the design effort on stiffness reduction and damping control. Five non-dimensional performance metrics are introduced that collapse the scattered literature into a single, scale-independent language for rigorous comparison across all absorber families: normalised half-absorption bandwidth ${\Pi }_1$, volume efficiency ${\Pi }_2$, integral absorption criterion ${\Pi }_3$ tied to the Rozanov causality bound, quality factor $Q=1/{\Pi }_1$, and frequency-thickness ratio ${\Pi }_4$. A two-degree-of-freedom acoustic–structural coupling model yields closed-form effective stiffness and damping, revealing how structural loss augments acoustic damping, how modal veering produces split absorption peaks, and how the anti-resonance frequency becomes a designable parameter. A critical distinction is drawn between mathematical negative stiffness (a fitting artefact) and physical negative stiffness via repulsive magnets, bistable elements, or negative-capacitance piezoelectric shunts, which genuinely reduces cavity stiffness, lowers resonance frequency, and widens bandwidth beyond the passive causality bound. The shunt electromechanical diaphragm further demonstrates $\alpha >0.9$ at nine tonal frequencies spanning three octaves without mechanical modification. Finally, embedding MSD equations and ${\Pi }_1$–${\Pi }_4$ bounds as hard physical priors in AI/LLM-assisted design frameworks is identified as the key step toward provably physically consistent absorber synthesis. Full article

To addressthe challenge of attitude control in quad-ducted fan unmanned aerial vehicles (UAVs) under coupled disturbances comprising thrust lag, ground effect and a composite wind field during near-ground flight and to mitigate the inherent trade-off between disturbance rejection and noise suppression in fixed-bandwidth extended state observers (ESOs), this paper proposes a robust attitude control method based on a Sigmoid law adaptive-bandwidth extended state observer (AB-ESO). An attitude dynamic model covering the above multi-source disturbances is established, with all uncertainties uniformly treated as lumped disturbances. An adaptive-bandwidth mechanism with filtering and rate-limiting modules is designed for smooth continuous bandwidth tuning. A composite control framework integrating disturbance feedforward, lag compensation and attitude feedback is constructed, and the uniform ultimate boundedness of the closed-loop system is proved. Comparative simulations are conducted against six baseline controllers, including a cascade proportional–integral–derivative (PID) controller, fixed-bandwidth ESOs, incremental nonlinear dynamic inversion (INDI), fast terminal sliding mode control (FTSMC) and a time-varying bandwidth ESO, in a near-ground composite wind scenario. Results show that the proposed method achieves improved comprehensive performance: the three-axis average tracking root mean square error (RMSE) is approximately 72% lower than of the PID controller and 15.8% lower than that of the high-bandwidth ESO, and the control output total variation is reduced by about 27.8%. Monte Carlo verification with 100 random turbulence groups further validates the strong statistical robustness of the proposed method. All validations in this work are based on numerical simulations. This study provides a technical reference for high-precision control of ducted fan UAVs in near-ground environments. Full article

Underwater wireless optical communication (UWOC) has emerged as a key technology for high-speed, low-latency data transmission in aquatic environments, enabling applications in autonomous underwater vehicles (AUVs), remotely operated vehicles (ROVs), subsea sensor networks, and the Internet of Underwater Things (IoUT). This paper reviews photodetector technologies that shape UWOC system performance, covering both mature and emerging detector classes. We discuss the operating principles, key parameters, and practical trade-offs of photomultiplier tubes (PMTs), p-i-n photodiodes (PINs), avalanche photodiodes (APDs), single-photon avalanche diodes (SPADs), and silicon photomultipliers (SiPMs/MPPCs). We also present emerging photodetector technologies, including perovskite-based structures, SiC photoelectrochemical devices, scintillating optical fibers, and photovoltaic solar cells. A comparative analysis of reported UWOC experiments reveals a clear sensitivity–bandwidth trade-off among detector technologies: PIN-based receivers achieve the highest data rates (up to 25 Gbps) but are generally restricted to short-range links, whereas SPAD- and SiPM-based receivers provide sensitivities below −80 dBm and support transmission distances exceeding 200 m, at the cost of moderate data rates. The findings indicate that SiPM/MPPC arrays currently offer the most promising compromise between sensitivity and data rate for long-range UWOC applications. Full article

This paper presents a fully integrated three-stage stacked power amplifier (PA) operating in the K-/Ka-band, fabricated using a TSMC 65 nm CMOS process for fifth-generation (5G) millimeter-wave applications. To improve the output power without sacrificing reliability, a two-stacked-FET configuration is adopted in the power stage, which distributes the voltage stress across series-connected transistors and permits a higher supply voltage than a conventional cascode structure. A differential topology with capacitive neutralization is employed in both the common-source driver stage and the stacked power stage. The neutralization capacitance is determined from a stability- and gain-oriented analysis based on the Rollett stability factor (K) and the maximum available gain (MAG), which clarifies the trade-off between feedback cancelation and wideband gain flatness over the operating band. The fabricated PA occupies a total chip area of 0.51 mm², including pads, with a core area of 0.18 mm². The measurement results demonstrate a measured 3 dB bandwidth of 6.7 GHz from 24.6 GHz to 31.3 GHz, a peak small-signal gain of 32.9 dB, a saturated output power of 20.8 dBm, an output 1 dB compression point of 16 dBm, and a peak power-added efficiency of 20.4% at 28.5 GHz. Over the 26–30 GHz range in which the large-signal characterization was performed, the saturated output power varies by less than 1 dB and the efficiency by less than 2 percentage points, indicating that the combination of capacitive neutralization and stacked-FET operation provides a compact solution for wideband K-/Ka-band CMOS power amplification. Full article

The increasing frequency and intensity of wildfires has created an urgent demand for scalable and autonomous wildfire response systems. While recent advances in multi-agent reinforcement learning (MARL) have demonstrated promise for collaborative uncrewed aerial vehicle (UAV)-based wildfire suppression, most existing approaches rely on simplified fire propagation dynamics and highly centralised learning architectures that are difficult to deploy in realistic operational settings. This paper presents a decentralised MARL framework for wildfire suppression that combines stochastic wildfire propagation, wind-driven spread dynamics, and communication-aware multi-agent coordination. The proposed framework extends an existing probabilistic wildfire environment through the incorporation of wind speed and directional effects, producing highly asymmetric and stochastic wildfire behaviour that more closely resembles real wildfire propagation. A decentralised Deep Q-Network (DQN) architecture is then introduced in which UAV agents learn independently through individual replay buffers. To mitigate the sparse-learning challenges introduced by decentralisation, selective experience sharing based on the SUPER algorithm is incorporated, enabling agents to exchange only high-value experiences under realistic communication constraints. Experimental results demonstrate that selective communication significantly improves containment performance and learning efficiency while preserving decentralised execution. The work highlights both the feasibility and challenges of realistic UAV swarm coordination for wildfire suppression, particularly the trade-offs between communication bandwidth, environmental stochasticity, and collaborative performance. Full article

This paper presents a linear mathematical model of bass guitar string vibration with experimentally identified, frequency- and fret-dependent modal damping, aimed at high-fidelity generative sound synthesis. To identify the string damping parameters across various frets and configurations, an experimental framework was developed to benchmark four structural identification methods: half-power bandwidth, I. Yoshida’s method, Discrete Fourier Transform Interpolation, and Hilbert-transform envelope approximation. Experiments were systematically conducted on Cort C4H, Ibanez RB 630, and Yamaha bass guitars. Based on the extracted parameter space, two audio generation strategies are formulated: a spectrum-driven harmonic reconstruction method (Method 1) and a physical modeling approach utilizing spatial wave equations (Method 2). The proposed linear approximation framework effectively captures the inverse relationship between the damping factor and fret numbers specifically on the E-string, while mapping linear increases on the G and D-strings. Quantitative verification using Sobolev norm differences demonstrates good agreement between the synthesized and original signals for the spectrum-driven method (Q = 0.031–0.057) and moderate agreement for the physics-based wave equation method (Q = 0.058–0.153). This reflects a trade-off in which the former achieves tighter spectral convergence, while the latter better preserves the physical, time-domain waveform structure. As both synthesis strategies are closed-form and computationally lightweight, the model is suitable for real-time implementation and the dynamic control of playing techniques (e.g., plucking location and, in principle, slap-type excitation), without relying on heavy, multi-gigabyte audio sample libraries. Full article

Transimpedance amplifiers (TIAs) are the critical components for current-to-voltage interfaces in optical receivers, LiDAR front-ends, biomedical sensors, and unconventional applications such as magnetic-resonance receiver-coil arrays and wide-bandgap ultraviolet detectors, and their CMOS design is governed by a fundamental gain-bandwidth-noise trade-off whose structure is rarely made explicit. This review introduces a unifying framework rooted in three explicit assumptions underlying the classical shunt-feedback TIA limit: a single-pole core amplifier (A1), a resistive feedback element (A2), and the full input capacitance loading the feedback summing node (A3). Relaxing one or more of these assumptions is shown to be the common structural thread behind every class of bandwidth or noise enhancement in the recent literature, and all surveyed architectures are organized into a six-tier taxonomy, from Tier 0 designs operating within the classical limit to Tier 5 topologies that bypass all three assumptions simultaneously. This taxonomy is supplemented by an orthogonal configurability axis spanning single- and dual-control reconfigurable, variable-gain, and dynamic-range-extension designs. We further show that stability is not removed by these relaxations but migrates with the tier, from the global phase margin of the classical loop to a local regulating loop, a group-delay-flatness constraint, an input-passivity condition, or a multi-loop interaction, so that each architecture carries a predictable stability locus. The taxonomy is cross-referenced with application domains and closed-form noise-floor boundary plots parametrized by input capacitance and amplifier gain-bandwidth product, and with the CMOS technology landscape, where we argue that the most advanced node is not universally optimal and that node and topology act as complementary rather than competing levers. A single consistent figure of merit, applied uniformly to a representative set of CMOS realizations from 0.6 $\mathsf{\mu }$m to 16 nm FinFET, shows no monotonic improvement with publication year or node and is presented as a diagnostic indicator rather than an absolute ranking. The review closes with an outlook on 200 Gb/s/lane links, wide-bandgap sensor integration, and the FinFET-to-gate-all-around device transition. Full article

This paper investigates frequency manoeuvring as a method to improve the resilience of unmanned aerial vehicle (UAV) networks operating in contested electromagnetic environments. The study considers scenarios in which the network initially operates on a single channel and is then exposed to intentional jamming or unintentional interference affecting the primary channel, adjacent channels or a wider frequency range. Several response policies are compared, including no channel change, immediate switching after quality degradation is detected, delayed switching after a defined loss-of-connectivity interval, and periodic frequency hopping. In addition to channel switching, the analysis also considers changes in channel bandwidth, comparing narrower channels with lower throughput but potentially higher resistance to interference against wider channels with greater capacity but increased susceptibility to disruption. The evaluation includes the switching cost, which is modelled as temporary packet loss, additional delay and jitter during reconfiguration. Performance is assessed using the packet delivery ratio, latency, jitter, packet loss and communication continuity. The main objective is to identify the interference conditions under which frequency manoeuvring becomes operationally beneficial and to determine which policy offers the best trade-off between resilience and communication performance. In quantitative terms, immediate switching under environmental interference achieved a PDR of 0.961 and a mean latency of 123.6 ms compared with a PDR of 0.946 and a mean latency of 138.7 ms for fixed-channel operation. Manoeuvring gave a substantial 12.2-percentage-point PDR gain under jamming (periodic hopping: 0.780 vs. 0.658) and a 6.7-percentage-point gain under combined interference (0.674 vs. 0.607). These results indicate that manoeuvring is most worthwhile once interference is persistent and channel-focused rather than purely environmental. Full article

Diesel generator sets are key frequency-supporting units in islanded microgrids and shipboard power systems, where rapid speed recovery under abrupt load variations is essential for maintaining power quality. However, conventional linear active disturbance rejection control (LADRC) is limited by the disturbance-estimation and noise-amplification trade-off of a single observer, while fixed parameters restrict its adaptability under varying operating conditions. To address these limitations, this paper proposes an RBF neural-network-optimized cascaded LADRC method, termed RBF-CLADRC. A mechanism-based torque balance model is first established, with uncertain mechanical coupling, friction losses, and load variations lumped into the total disturbance. A residual-disturbance cascaded observer is then constructed, in which the first linear extended state observer estimates the total disturbance and the second further reconstructs the residual estimation error. Unlike conventional ML-based ADRC methods that directly tune multiple gains, the proposed RBFNN adjusts only a common controller bandwidth within a prescribed interval, while all observer and feedback gains are generated through predefined analytical relationships. This low-dimensional adaptation preserves coordinated gain variation, reduces online computational complexity, and facilitates real-time implementation. Lyapunov analysis shows that the observer and tracking errors are uniformly ultimately bounded under bounded disturbance rates and converge exponentially for constant disturbances. Finally, comparative simulations in MATLAB/Simulink demonstrate that the proposed method achieves better dynamic response and disturbance-rejection performance than conventional LADRC and other benchmark controllers. Full article

To address the challenge of simultaneously achieving broadband absorption, multi-mechanism tunability, and angular stability in terahertz multifunctional devices, this paper proposes a MoS₂/VO₂ composite terahertz metamaterial absorber based on an asymmetric multi-nested C-shaped structure. The device adopts a three-layer configuration consisting of a MoS₂/VO₂ composite plane–SiO₂ dielectric–Au reflector layer. Unlike conventional symmetric structures, which are limited by selection rules and symmetry-protected dark modes that hinder the excitation of higher-order resonances, this design effectively breaks structural symmetry protection through geometric asymmetry. This induces strong mode hybridization between originally orthogonal dark and bright modes, enabling broadband high absorption exceeding 96.7% across the 1.88–3.52 THz frequency range (61% RBW). Notably, the device demonstrates synergistic tuning advantages: the macroscopic on/off switching of broadband absorption characteristics via the phase transition of VO₂, combined with fine blind-spot compensation and enhancement in absorption peaks using the electrical tunability of MoS₂. Furthermore, thanks to its sub-wavelength unit cell design, the structure maintains excellent performance stability over a wide incident angle range from 0° to 60°. This study reveals a synergistic enhancement mechanism combining the asymmetric unit cell and hybrid materials, providing a systematic physical solution for resolving the trade-off between bandwidth extension and dynamic reconfigurability. Full article

LCL-type grid-connected inverters face problems including complex modeling, sampled current distortion from harmonics and negative-sequence components, and reduced control accuracy of conventional deadbeat predictive current control (DPCC) due to its heavy reliance on precise system parameters. To solve these issues, this paper proposes a model-free DPCC (MF-DPCC) using an adaptive-gain extended state observer (AGESO). Firstly, an ultra-local model is established to avoid dependence on accurate mathematical models. Secondly, an AGESO is designed to overcome conventional ESO drawbacks (initial differential peaking, inflexible bandwidth tuning, and tracking-noise immunity trade-off) by adopting adaptive gains to real-time estimate the ultra-local model’s lumped disturbance and state variables. Finally, a double second-order generalized integrator (DSOGI) purifies sampled currents and extracts fundamental positive-sequence components, reducing harmonic disturbance on the AGESO, allowing for higher bandwidth operation without excessive noise amplification, and indirectly enhancing resonance suppression by including LCL resonance-induced disturbance in the lumped term. Full article

Software-defined radio (SDR) has become a critical technology in modern communications, offering the flexibility and programmability required for dynamic signal processing. As the demand for high-performance SDR systems continues to increase, selecting an appropriate hardware platform has become increasingly important. This paper investigates the implementation of a state-of-the-art Chessboard-based Automatic Modulation Classification (CAMC) algorithm on a Field-Programmable Gate Array (FPGA)-based RFSoC platform, exploring various parallel datapath architectures and their impact on performance. The relationships between maximum operating frequency, bandwidth, and resource consumption are evaluated as the number of parallel instances and peripheral configurations are varied. The results show that the parallel CAMC design, incorporating all required peripherals, achieves a bandwidth of 29.0 GBps while consuming 82.31% of lookup tables (LUTs) and 50.67% of flip-flops (FFs), whereas the datapath-only design achieves a bandwidth of 24.8 GBps with significantly lower resource consumption. In addition, as the number of parallel instances increases, the number of High-Performance (HP) ports or master ports increases, the maximum operating frequency decreases, and bandwidth growth gradually saturates. The presented implementation study and open-source hardware designs provide practical reference architectures for evaluating the trade-offs between bandwidth, performance, and resource utilisation in FPGA-based SDR systems. Full article